THE EXCEPTIONAL SOFT X-RAY HALO OF THE GALAXY MERGER NGC 6240

Besides the ordinary image reprocessing, further corrections are necessary in the study of extended X-ray sources to convert counts into physical units, properly taking into account the dependence of the collecting area on energy and position, as well as the effective exposure in the different regions of the detector. With this aim, we have first built an instrument map using the ciao tool mkinstmap, and then generated an exposure map (see Davis 2001) for each observation with mkexpmap. The input spectral weights were calculated by extracting a preliminary halo spectrum from the r = 20''–80'' annulus centered on the brightest, nuclear hard X-ray source, and modeling it with a phenomenological two-component thermal model. The individual maps were then reprojected and combined to match the merged image.

In order to assess the maximum halo extent along the various directions, we initially focused our attention on the 0.7–1.1 keV band, which represents the optimal compromise between the requirements of both high S/N and total net counts. At the standard soft X-ray energies of 0.3–2 keV, in fact, the source counts amount to ∼10,000, but this is only 59% of the whole number of events collected over that range. By restricting to 0.7–1.1 keV, instead, the same fraction rises to 77% while retaining ∼6000 net counts. This improves the sensitivity at low surface brightness (see Figure 1).

Zoom In Zoom Out Reset image size

In ObsID 12713, the nuclear region of NGC 6240 lies ∼2' from the western edge of the S3 chip. Even after applying the exposure correction, this puts a rough empirical limit on the area that can be effectively explored for the presence of diffuse emission (at least in the east–west direction), as the large exposure disparity (≈182/37) would affect the significance of any detection. This notwithstanding, we have adaptively smoothed the 0.7–1.1 keV image to emphasize the faint extended features while preserving the local details with the highest S/N, using csmooth (Ebeling et al. 2006). This tool convolves the raw image with increasing Gaussian kernels; we set a nominal significance between 2.5σ and 5σ over the local background under each kernel, and imposed a maximum kernel size of 10 pixels to prevent any oversmoothing. The same smoothing scales were applied to the exposure map, which was eventually used to normalize the source image.

The final, exposure-corrected image is shown in Figure 1, where also the detection contours at the 2σ and 3σ confidence level above the average background have been computed. The latter curve is taken to represent the outer boundary hereafter, although its mean radius of ∼80'' can still be regarded as conservative measure with some respect (e.g., the overly fine spatial resolution; see below). According to this estimate, the X-ray halo of NGC 6240 appears to have a somewhat distorted, elongated, diamond-like shape. The directions of minimum and maximum extension intersect to the southeast of the nucleus and are almost exactly perpendicular, corresponding to position angles of ∼16° and 107°, respectively. The physical size at full length is then ∼110 × 80 kpc (225'' × 162''). In Figure 2 we draw the logarithmic contours starting at the fiducial detection limit, with a surface brightness cutoff that brings out the structures within the halo. As already pointed out by Bush et al. (2008), the most conspicuous feature is the stretched, cross-like pattern characterizing the central r ∼ 1' region, whose arms generate a bicone with an aperture angle of ∼130°. This is even more evident below in Figure 4, where the emission clumpiness is visibly enhanced. Notably, the axis of symmetry of this bicone is virtually normal to the plane of the optical disk, and its apex is coincident with the hard X-ray source. This is all commensurate with a galactic outflow arising from the nuclear starburst similar to that observed in M82 (Griffiths et al. 2000), and possibly sweeping across a pre-existing halo medium. The physical scales involved, however, are rather challenging (see Section 5).

Zoom In Zoom Out Reset image size

Compared to the oblong structure above, at larger distance (r > 1' ∼ 30 kpc) and out to the detection boundary the hot gas distribution tends to recover a rounder shape. In spite of the overall slight asymmetry, a radial analysis can provide valuable information on the system. We have then computed the surface brightness profile of NGC 6240 up to a galactocentric distance of 128'', sampling this region with successive, 4'' wide annuli. The surrounding field is quite crowded, hence we employed the wavelet-based algorithm wavdetect (Freeman et al. 2002) to isolate any contaminating point-like sources, three of which lie just within the outer border. As their non-negligible contribution to the respective annuli might substantially alter the slope of the faint emission tail and the detection significance, we excluded a circle of 4 pixel radius around each of these sources.

The resulting background-subtracted surface brightness profile was fitted with a two-component β-model, which is a convenient analytical expression widely used in the study of galaxies and clusters, defined as

$$\Sigma (r)=\Sigma _0 \left[ 1+\left(\frac{r}{r_{\mathrm{s}}} \right)^2 \right]^{-3\beta +0.5},$$

where Σ₀ is the central surface brightness, r_s is the scale radius of the gas distribution, and β is the power-law index. With β = 0.5, for instance, Σ∝r⁻² at r ≫ r_s. Based on the shorter ObsID 1590 only, Huo et al. (2004) already suggested that a double β-model better accounts for the radial profile of NGC 6240 than the simpler form above. Besides the confirmation of these earlier findings, we are now able to show that the core and halo components can be completely disentangled (Table 2; Figure 3). In particular, the extended, slowly declining component equals its compact, steeply falling counterpart ∼7 kpc away from the center, and becomes dominant straight beyond. We have consequently assumed r = 15'' as the inner boundary of the halo in our later spectral analysis. This is consistent with a rough estimate from visual inspection of the X-ray morphology.

Zoom In Zoom Out Reset image size

According to the radial surface brightness profile, diffuse emission associated with NGC 6240 is detected at the 3σ confidence level out to a galactocentric distance of ∼100'' (50 kpc). Since counts are now integrated over a given annulus, this measure naturally overrides the single pixel-based 3σ contour formerly supplied, and is our best approximation to the actual average extension of the halo.

The clumpiness of the morphological appearance suggested by Figure 1, as opposed to a possible flatness, can be itself indicative of the nature of the halo. In order to assess the degree of complexity, we have performed the adaptive smoothing with the alternative ciao tool dmimgadapt, which is also available within DS9.¹⁴ Differently from csmooth, in this case the underlying algorithm returns a superior resolution of the bright small-scale structures. Indeed, within csmooth the kernel significance is computed with respect to the local background. This magnifies the faint, diffuse component close to the halo border, yet inner structures may be not significant enough with respect to the neighboring area. Such features are then smoothed at the widest scale allowed, and any clumpiness is lost. On the contrary, dmimgadapt is based on the raw number of counts. We then switched to the 0.5–1.5 keV image, due to its ∼13,000 total counts, 71% of which are associated with the source. We adopted the same maximum size as before (10 pixels), and required at least 16 counts under each kernel. Given the background expected, this corresponds to a minimum significance of 1σ, growing up to 3.5σ for the finest (i.e., <4'' wide) structures. The results are illustrated in Figure 4, where knots and filaments are clearly visible throughout the physical scales across the halo.

Zoom In Zoom Out Reset image size

We further addressed this issue from a quantitative point of view by selecting a sample direction across the halo, off the nuclear region but intersecting several apparent, local intensity peaks. We extracted the number of counts along a 200'' × 4'' stripe, divided in 50 square cells. Figure 5 shows the tight correlation between the raw and the smoothed image; the actual difference of count density substantiates the statistical significance of the small-scale structures. In conclusion, these features are not consistent with a fully relaxed environment in nearly thermal and hydrostatic equilibrium, but rather they are reminiscent of the loops, bubbles, and cavities seen in the central r < 7 kpc region of NGC 6240, where the shock-heated gas is subject to violent starburst-driven winds and, possibly, to massive AGN feedback (Lira et al. 2002; Komossa et al. 2003; Feruglio et al. 2013). A spatially resolved spectral analysis can then provide crucial information to constrain the properties of the hot gas in the halo, and to shed light on the origin and evolution of the whole system.

Zoom In Zoom Out Reset image size

The preliminary spectrum extracted during the calculation of the exposure map hinted at soft X-ray line emission within the halo. Using the XMM-Newton spectrum (70'' extraction radius) presented by Boller et al. (2003) as a reference, we were able to identify the Fe-L complex, possibly blended with O vii–viii and Ne ix–x (0.55–1.12 keV), and also Mg xi–xii (1.24–1.5 keV) and Si xiii–xiv lines (1.7–1.93 keV). Following Baldi et al. (2006a) and Wang et al. (2009), we generated three narrowband images to encompass the features listed above, in order to obtain a visual map of line strengths and metal abundances in the hot, diffuse interstellar medium. For the continuum subtraction, another image was extracted from the joint 0.4–0.55, 1.12–1.24, and 1.5–1.7 keV energy range. All the line and continuum bands were selected to achieve an optimal compromise between line strength and continuum contamination. No culling was applied for the few point sources. The O–Fe–Ne image was adaptively smoothed (with csmooth) as described earlier. Due to its highest S/N and largest number of line counts (∼4000), the same smoothing scales were adopted for the other images. The background level appropriate to each energy range was then subtracted, while the continuum contribution to each line image was determined by fitting the halo spectrum with an absorbed bremsstrahlung model, after excluding the strong emission features. The resulting scaling factors (∼2.2, 0.4, and 0.1 for the O–Fe–Ne, Mg, and Si bands, respectively) were assumed as relative weights in the final continuum subtraction.

The accuracy of this procedure is corroborated by the little spectral variations across the halo (see below), which allowed us to employ the average continuum. Indeed, this is not the case for the core region, where the source emission is much more complex and position dependent. As a result, most of the structures found in the line-strength maps within the central 15'' have no physical meaning. For this reason, in Figure 6 we only show the lowest significance contours of the soft X-ray emission lines under examination. The comparison with the archival Hubble Space Telescope ACS/WFC F814W image reveals that metals are spread out far beyond the optical limits of the galaxy. Not surprisingly, given the energy band and the line intensity, the O–Fe–Ne morphology turns out to be very similar to that illustrated in Figure 2. Once the intrinsic difference in S/N is taken into account, also the spatial distribution of Mg and Si line emission displays a close overlap. In particular, the prominent cross-like shape is clearly recovered in the 3σ Mg contours. This suggests that the halo of NGC 6240 has already experienced a significant, nearly uniform metal enrichment out to very large distances.

Zoom In Zoom Out Reset image size

The radial surface brightness profile of Figure 3 indicates a sharp separation at r = 15'' ≃ 7.5 kpc between the core and the halo components. Starting at this inner boundary, we followed the 3σ contours of the 0.5–1.5 keV image to define the width of the extraction region for the spectral analysis. Due to the slight asymmetry, the outer border is not fixed, but varies from 75'' to 100'' over six azimuthal sectors (labeled S1–S6; see Figure 7). Qualitatively then, the full halo (FH) region has the shape of an irregular windmill, and contains ∼11,300 net counts over the entire 0.3–8 keV energy range. With no loss of statistical information, we restricted the range to 0.4–2.5 keV (∼10,700 net counts) for the spectral fitting.

Zoom In Zoom Out Reset image size

Ten point-like, soft X-ray sources were identified within the halo extraction boundaries by running wavdetect over the selected band. Each of them yields a few tens of counts at most, for an aggregate contribution to the extended emission of less than 3%. Given their minor weight, these point sources were not excluded. Their cumulative spectrum is well reproduced by an absorbed (N_H ∼ 10²¹ cm⁻²) power law with photon index set to 1.8, whose intrinsic 0.3–8 keV luminosity is <3 × 10⁴⁰ erg s⁻¹ at the distance of NGC 6240. Owing to their location, six are allegedly background AGNs (but no optical counterpart is known), while the remaining four, if local to the galaxy, would definitely be ultraluminous X-ray sources (ULXs; e.g., Zezas & Fabbiano 2002). The observed spectral shape is also consistent with the unresolved populations of X-ray active objects, including high- and low-mass X-ray binaries, active binaries, and cataclysmic variables (e.g., Fabbiano 2006; Boroson et al. 2011). These sources are associated with the stellar population, hence their contribution to the halo is expected to be negligible. We accounted for any residual emission from the latter classes (as well as for the resolved point sources) by keeping the Γ = 1.8 power-law component in our model.

The bulk of the soft X-ray emission in the halo is presumed to arise from diffuse hot gas, and was then modeled as a thermal spectrum through the vapec code (Smith et al. 2001), which makes use of the AtomDB v2.0.1 atomic database.¹⁵ The gas emission measure (EM) can be expressed as a function of the vapec normalization $\mathcal {N}_{\mathrm{v}}$ through the relation ${\mathrm{EM}}=\int n_e n_{\mathrm{H}} {d}V=4\pi [D_{\mathrm{A}}(1+z)]^2 \times 10^{14} \mathcal {N}_{\mathrm{v}}$, where n_e and n_H are the electron and hydrogen densities, and D_A is the angular diameter distance to NGC 6240 (102 Mpc). We first considered a single-temperature plasma component, allowing for local (i.e., at the redshift of the source) absorption in addition to the Galactic column density (N_H = 4.87 × 10²⁰ cm⁻²; Kalberla et al. 2005). The model form is then expressed within xspec as wabs*zwabs*(vapec+powerlaw). As mentioned, the halo region provides ∼10,700 counts, almost 90% of which are detected at 0.5–1.5 keV, with an improvement by a factor of ∼4 with respect to ObsID 1590 alone. This enabled us to assess for the first time individual abundances for iron and each of the main α-elements (O, Ne, Mg, Si). For all the other elements, we have adopted solar abundances (from Anders & Grevesse 1989). Even if this standard practice might lead to some inaccuracies, it is not really critical here, as the overall contribution of the frozen elements to the best fit is marginal (<20%). The only possible exception is represented by nickel L-shell emission (∼6%; see later in the text). We refer to Kim (2012) for a review of all the systematic uncertainties affecting the measure of elemental abundances through X-ray spectral fitting to CCD data.

The spectrum of the halo region is plotted in Figure 8, while the best-fitting parameters are listed in Table 3. Model A applies to the rebinned spectrum, for which the χ² statistic is assumed, while model B is relative to the ungrouped data, treated with C-stat (see the Appendix for an overview). The results are in full agreement. Interestingly, the amplitude of the power-law continuum and the column density of the local absorber have a null best-fit value, implying that both components are not required, and that the diffuse, soft X-ray emission can thus be accounted for within a simple thermal scenario (χ²_ν ∼ 1.14). The upper limit to the 0.3–8 keV power-law luminosity is 4.1 × 10⁴⁰ erg s⁻¹, encompassing the contribution estimated above for the point sources. On the other hand, NGC 6240 is known to host a huge amount of cold gas, but this is concentrated in the nuclear regions (Baan et al. 2007). In model C we fitted the six halo subregions S1–S6 simultaneously with a wabs*vapec model, where the thermal emission is modified by Galactic absorption only. Individual temperatures and metal abundances are tied, and are found to be in good agreement with the fiducial values.

Zoom In Zoom Out Reset image size

The main result is a general metal underabundance (by a factor of ∼2 for α-elements) with respect to solar (Table 3). Absolute abundances should be taken with some caution, since they strongly depend on the determination of the continuum, which, in turn, is affected by unknown factors such as absorption, non-thermal components, and multi-temperature gas emission (e.g., Kim 2012). We then tried introducing a second thermal component to probe the possible presence in the halo of multi-temperature gas. Due to the heavy degeneracy between the column density and the temperature of the additional gas phase, which favors a statistically equivalent (but physically unacceptable) solution,¹⁶ we have dropped both local absorption and power-law continuum, adopting a wabs*(vapec+vapec) form (model D). The goodness of fit is significantly increased (χ²_ν ∼ 1.02), and calls for a mixed plasma with kT₁ ∼ 0.9 keV and kT₂ ∼ 0.4 keV, while the two emission measures are comparable. The results are illustrated in Figure 8, and summarized in Table 3. Notably, there is no substantial change in the abundances of iron and of the main α-elements, which were still allowed to be different fractions of solar, but were tied between the two gas components since they cannot be constrained separately. In model E, instead, abundance was fixed to 0.1 solar for all the elements in the low-temperature component. Both kT₁ and kT₂ decrease by ∼0.15 keV, yet the fit is further improved (Δχ² ≃ −2.7 with the same degrees of freedom), supporting the presence of complex physical conditions in the halo of NGC 6240. This notwithstanding, the single-temperature model clearly represents a sound approximation to the data, and provides useful parameters for the purpose of this work.

As a further test, we considered the possibility that collisional ionization equilibrium (the underlying assumption of the vapec code) is not met by the shock-heated gas in the halo. NGC 6240 is undergoing a violent, merger-induced starburst phase, which is responsible for most of the system's ∼10¹² L_☉ IR luminosity (Lutz et al. 2003; Nardini et al. 2009). The last major burst of star formation is estimated to have occurred ∼20 Myr ago (Tecza et al. 2000). Whether a pre-existing halo medium is swept up and shock-heated by a starburst-driven wind, or shocks are taking place inside the wind material itself, a low-density (n_e ∼ 10⁻³ cm⁻³) gas may have not recovered ionization equilibrium conditions yet. The typical, density-weighted timescale for plasma to reach equilibrium is τ_eq = n_et > 10¹² cm⁻³ s (Masai 1994; Smith & Hughes 2010). We then switched from vapec to the non-equilibrium model vnei (Borkowski et al. 2001), fixing the ionization timescale parameter at τ_eq = 10¹¹ cm⁻³ s. This returned a much poorer spectral description (χ²_ν ∼ 2), hence we did not pursue this interpretation any further. Incidentally, by taking τ_eq = 3 × 10¹² cm⁻³ s we obtained χ²_ν ∼ 1.3, proving that vnei, if unconstrained, tends to converge toward an equilibrium solution.

Finally, we investigated the nickel L-shell issue in more detail, by repeating the spectral fits (model A) with different values of nickel abundance. This was in turn fixed to half-solar, solar, and twice-solar, tied to iron, and left free. This check delivers a nearly evenly spaced grid of Z_Ni values, from which we can draw some qualitative considerations in spite of the large uncertainties involved. Although the corresponding best fit is still broadly acceptable (null hypothesis probability of ∼11%), the assumption of Z_Ni ≡ Z_Fe ≃ 0.15 results in a significant worsening on statistical grounds (Δχ² ≃ 5.1). Indeed, nearly solar values of Z_Ni are definitely preferred. Equivalent results were found with C-stat on the ungrouped spectra. Once an obvious nickel underabundance is ruled out, most elements are virtually unaffected by the exact entry for Z_Ni, due to the substantial difference from nickel in either total contribution (Fe) or energy (O, Mg, Si). Conversely, neon abundance varies systematically with Z_Ni, decreasing by a factor of ∼3 as nickel changes from half to twice solar, as the two elements are very similar in both flux and energy. Thawing Z_Ni has a negligible effect on the goodness of fit (Δχ² ≃ −0.6). An F-test gives a probability of chance improvement of 48%. The best-fit value is Z_Ni ∼ 1.4, but poorly constrained in the 0.5–2.5 range. We have therefore kept nickel abundance frozen to solar for all the subsequent analysis.

Given the fairly large amount of total net counts available, the next step is to embark upon a spatially resolved analysis aimed at revealing any temperature and/or metallicity gradient within different zones of the halo. We first searched for some trend in the radial direction, and selected two annular regions corresponding to the inner (IH; r = 15''–36'') and outer (OH; r = 36''–80'') halo (see Figure 7). The IH/OH transition radius was adaptively chosen to ensure an equal number of source counts (∼4500) in the 0.5–1.5 keV energy band. Due to the different areas (by a factor of ∼5), however, the S/N in the two subregions is quite different (see Table 4). In the following, we systematically refer to the analysis of the ungrouped spectra, which preserve the intrinsic ACIS-S energy resolution. This entails sounder measures and more reliable comparisons among elemental abundances, as Mg and Si lines fall in a rather noisy spectral range. Even the O–Fe–Ne blend can be affected by a conservative rebinning like the one adopted to employ the χ² minimization.

Table 4. Summary of the Different Extraction Regions

Reg	$\mathcal {C}_{0.5\hbox{--}1.5 \,{\mathrm{keV}}}$	$\mathcal {C}_{0.4\hbox{--}2.5 \,{\mathrm{keV}}}$	$\mathcal {F}_{s/t}$	Reg	$\mathcal {C}_{0.5\hbox{--}1.5 \,{\mathrm{keV}}}$	$\mathcal {C}_{0.4\hbox{--}2.5 \,{\mathrm{keV}}}$	$\mathcal {F}_{s/t}$
FH	9463 ± 147	10675 ± 173	63.7	IH	4560 ± 72	5176 ± 79	86.2
S1	1577 ± 52	1742 ± 60	58.2	OH	4526 ± 103	5084 ± 120	56.6
S2	1575 ± 48	1777 ± 54	67.3	H1	1837 ± 44	2108 ± 48	91.6
S3	1583 ± 47	1775 ± 52	70.5	H2	1802 ± 45	2044 ± 50	84.8
S4	1579 ± 53	1775 ± 61	57.4	H3	1826 ± 48	2032 ± 53	76.7
S5	1581 ± 47	1788 ± 53	69.3	H4	1816 ± 52	2030 ± 59	66.6
S6	1568 ± 50	1820 ± 57	63.3	H5	1804 ± 70	2046 ± 83	44.7

Note. $\mathcal {C}$: net counts over the reference energy band (0.5–1.5 keV) and over the full spectral range (0.4–2.5 keV); $\mathcal {F}_{s/t}$: fraction of source to total counts at 0.4–2.5 keV, in percent.

Download table as:  ASCIITypeset image

We have then used model B for the IH and OH fits. A marginal power-law component is possibly present in the IH region, consistent with the luminosity of the point sources that fall in that field. The upper limit to the total power-law contribution is in good agreement with the previous estimates in Table 3. The two IH and OH plasma temperatures are identical to the fiducial FH value, and again no local absorber is required. Ultimately, the two spectra are remarkably alike (Figure 9). The most significant comparison concerns abundances, as illustrated in Figure 10. In spite of the overlapping error bars, for each element the best-fitting OH values are systematically lower than the IH ones, suggesting a slight metallicity drop at larger distances. By disentangling the core and halo spectra, a similar result had been already found by Huo et al. (2004), who used a simplified model to fit the ObsID 1590 data (see also Grimes et al. 2005). Although negligible when considering the single element, the overall effect is significant at the 2σ level. As already mentioned, however, absolute abundances can be somewhat misleading. Relative abundances (i.e., abundance ratios between two elements) are usually more reliable, as the prominence of the different lines is altered in a similar way by any incorrect subtraction of the underlying emission. The decreasing trend is not seen in the abundance ratios. This can be explained in either of the following ways: (1) the observed metallicity drop is artificial, likely induced by an inaccurate determination of the continuum; or (2) the negative radial gradient has a physical origin, possibly related to the dilution of a metal-enriched gas outflow. Both interpretations are consistent with the apparent behavior of absolute and relative abundances.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Although the IH/OH analysis revealed no clear trend for either temperature or metallicity, we proceeded with a finer annular sampling of the halo. This is desirable in order to obtain a radial profile of the gas density as a function of the emission measures (see later). Following the same adaptive criterion, we selected five regions within r = 15''–80'' (H1–H5, with intermediate radii of 218, 308, 418, and 555), containing ∼1800 net counts each at 0.5–1.5 keV (Table 4). Model C was applied in the spectral analysis, with temperatures free and abundances tied to their average values. We also made use of the deprojection technique available through the xspec model component projct. Indeed, since we can only probe the footprint of the emitting source on the plane of the sky, a system with a rough spherical symmetry is subject to some degree of radial mixing: The emission observed in a given annulus also includes a contribution from material located at larger distances. In this case, however, deprojection does not result in any significant difference, confirming the suggested lack of strong radial gradients. Only emission measures experience some variations, but the correction involved for the gas density is still limited (<30%), and therefore has been disregarded.

As discussed above, radial variations, if any, are limited. A possible, partial explanation is that the integration over the polar angle smooths out most of the existing differences. From Figures 1 and 4, in fact, one can easily appreciate that the surface brightness profile depends on the direction along which it is evaluated. Several structures, outstanding for intensity and/or extent, are confined in narrow azimuthal sectors, so that the relative information might be lost in a purely radial analysis. The aforementioned azimuthal regions S1–S6 (Figure 7) were adaptively selected following the morphological appearance, and imposing that each key feature is sharply singled out, most notably the wings of the cross-like pattern. Each sector includes ∼1800 net counts at 0.4–2.5 keV, which represent ∼60%–70% of the total collected events (see Table 4). We then made use of two new models, F and G, both based on the simple wabs*vapec form. This choice is fully justified by our preliminary analysis, demonstrating that any additional spectral component is redundant, apart from a thermal one (which, however, cannot be explored over the individual zones).¹⁷

At first, we took the most general approach of dealing with each subregion independently (model F). The six spectra and their best fits are shown in Figure 11. The results for the individual gas temperatures and abundances are summarized in Table 5, and plotted in Figures 12 and 13. Evidence for a slightly hotter gas component is found in S5, and also abundance measures provisionally hint at sizable gradients. The error bars are quite large, though. With model G, we have therefore adopted an alternative strategy to reduce both the systematic and the statistical uncertainties on absolute and relative abundances. The basic idea is to take advantage of the fact that the extent of any variation across the halo has proven to be rather limited. This allowed us to extract some additional information by investigating the behavior of each of the key spectral parameters separately, with all the other variables tied (but not frozen) to their average value.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We created in turn six spatial maps for kT, and O, Mg, Ne, Si, Fe abundances by performing a simultaneous fit of the S1–S6 regions, with the single observable of interest free to vary in the different spectra and the other parameters (apart from the normalization) tied to a common value. According to this method, Table 5 should be read along rows for model F only. The physical information from model G is conveyed by columns, which correspond to a specific temperature or abundance map. Since no quantity is frozen, every fit delivers an independent set of average values for the collateral parameters. Not surprisingly, these are always coincident (typically within ∼5%–10%) with the results of model C, where temperatures and abundances are all tied over S1–S6, and from which the entries for emission measure and C-stat (evaluated on the single region of interest) have been taken.

Model G effectively returns smaller error bars, as illustrated in Figures 12 and 13. This result is not actually achieved at the expense of accuracy. The measure of individual abundances, in first approximation, relies on a specific, narrow energy range, whose continuum is now constrained over the full halo. The fact that C-stat undergoes just tiny increments with respect to model F, when assessed over the single regions, implies that the entire soft X-ray spectrum is well reproduced, not only the band of interest for each abundance map. This further confirms the reliability of our approach. The last sensitive issue to investigate is the possible degeneracy between iron abundance and emission measure, which can partly weaken the significance of any gradient. In spite of the low Z_Fe value, iron L-shell contribution is still dominant, and it is difficult to disentangle from the continuum at CCD resolution. In Figure 14 we show the Z_Fe–EM confidence contours for the two regions (S4 and S6) with the larger difference (2.8σ) in Z_Fe, proving that such effect is minimal. A gradient of iron abundance in the halo is then detected with a ∼2σ significance. With little dependence on the specific model (see Figure 13), a similar result possibly holds for magnesium, while for all the other α-elements error bars are still too large to draw any conclusion.

Zoom In Zoom Out Reset image size

The key physical properties (density, mass, pressure, thermal energy) of the X-ray emitting gas can be inferred from the results of the spectral analysis, provided that the volume of the halo is estimated. We assumed that the thickness of the system is comparable to its width. This is a reasonable approximation for a merger about to enter the final coalescence stage like NGC 6240. Moreover, the main galactic body in the optical is reminiscent of a distorted edge-on disk (Gerssen et al. 2004), implying a substantial span along the line of sight. A nearly face-on configuration (like in the case of the Antennae, NGC 4038/4039; Fabbiano et al. 2003) does not seem to apply here, and can be safely ruled out. Based on this conjecture, we computed the volume $\mathcal {V}$ of the halo as follows:

$$\mathcal {V} = \sum \frac{2 \theta _i}{3} R_i^3 \left(\frac{2 \mathcal {S}_i}{\theta _i R_i^2} \right)^{3/2} \simeq 1.9 \times \sum \mathcal {S}_i^{3/2} \theta _i^{-1/2},$$

where $\mathcal {S}_i$ is the surface derived for each of the six azimuthal regions S1–S6 from the 3σ contour of Figure 1, and R_i and θ_i are, respectively, the outer radius (in kpc) and the central angle (in radians) of the individual sectors. We eventually obtained that $\mathcal {V} \simeq 2.15 \times 10^5$ kpc³. In the detailed calculations, we have also taken the geometrical projection effects into account. This is mainly relevant to the volumes of the annular regions H1–H5, which strongly depend on the radius of the embedding sphere, taken to be 100'' (±20''). Such a correction is negligible (roughly 1%) on $\mathcal {V}$, also in light of the large systematic uncertainties.

It is actually the effective volume occupied by the gas that matters for the estimate of the physical properties. This is related to the geometrical volume through the expression $V_{\mathrm{eff}} = \eta \mathcal {V}$, where η encompasses both the filling factor of the emitting gas and any correction scale factor. Assuming n_e ∼ n_H, the emission measure can be written as ${\mathrm{EM}} \simeq n_e^2 \eta \mathcal {V}$, from which we extracted the electron density, and hence the gas pressure p = 2n_ekT, mass $M = n_e \eta \mathcal {V} m_p$, and thermal energy $E_{{\rm th}} = 3 n_e \eta \mathcal {V} kT$. The cooling time τ_c was estimated as the ratio between the thermal energy content and the 0.3–8 keV thermal luminosity, inferred from the spectral fits. Our results for the full halo and its different subregions are presented in Table 6, and their main implications are discussed in the next section.

Table 6. Global and Local Physical Properties of the Hot Gas

Reg	Mod	$\mathcal {V}$	ne	Mg	p	Eth	τc
FH	B	6.31	2.5+0.2−0.3	13.1+1.2−1.2	5.2+0.7−0.6	49+7−5	3.7+1.0−0.6
H1	C	0.54	3.6+0.5−0.4	1.7+0.2−0.2	8.3+1.1−1.1	6.7+0.9−0.9	2.6+0.3−0.4
H2	C	1.00	2.6+0.4−0.3	2.2+0.3−0.3	5.6+0.9−0.7	8.5+1.3−1.2	3.3+0.6−0.4
H3	C	1.63	2.1+0.3−0.3	2.8+0.4−0.3	4.4+0.7−0.7	10.7+1.6−1.6	4.2+0.7−0.6
H4	C	2.48	1.7+0.2−0.3	3.5+0.5−0.5	3.7+0.6−0.6	13.7+2.2−2.2	5.5+0.9−1.0
H5	C	2.95	1.5+0.4−0.3	3.8+0.7−0.9	3.0+0.8−0.7	13.2+2.9−3.5	5.3+1.2−1.4
S1	F	1.02	2.2+0.4−0.6	1.9+0.4−0.5	4.5+1.2−1.2	7.0+1.8−1.9	3.4+0.9−1.0
S2	F	1.06	2.8+0.3−0.4	2.5+0.3−0.3	5.7+1.2−1.0	9.0+1.9−1.6	3.9+0.8−0.7
S3	F	0.67	3.0+0.4−0.6	1.7+0.2−0.3	6.3+1.2−1.3	6.3+1.2−1.3	2.8+0.6−0.5
S4	F	1.25	2.1+0.4−0.5	2.3+0.4−0.6	4.4+1.0−1.1	8.2+2.0−2.1	3.8+1.0−1.0
S5	F	0.99	2.5+0.3−0.4	2.1+0.3−0.3	6.5+1.2−1.1	9.7+1.8−1.6	4.4+0.8−0.8
S6	F	1.32	2.4+0.3−0.5	2.6+0.4−0.5	5.0+1.3−1.1	9.9+2.5−2.2	4.3+1.1−1.0

Note. $\mathcal {V}$: estimated volume in 10⁶⁹ cm³; n_e: electron density in 10⁻³ cm⁻³; M_g: total mass in 10⁹ M_☉; p: pressure in 10⁻¹² dyne cm⁻²; E_th: thermal energy in 10⁵⁷ erg; τ_c: cooling time in Gyr.

Download table as:  ASCIITypeset image

All the physical quantities mentioned above contain the unknown filling factor η, which cannot be directly constrained. Hydrodynamic simulations suggest that the low-density, hot gas component (n_e ∼ 10⁻² cm⁻², kT ∼ 10⁷ K) in starburst-driven winds is volume-filling (η ∼ 0.7), even if its contribution to the soft X-ray emission may not be dominant (e.g., Strickland & Stevens 2000). Since the dependence from η is mild (n_e, p∝η^−1/2, while M_g, E_th, τ_c∝η^1/2), the estimates given in Table 6 correspond to η = 1, keeping in mind that the actual values can differ by a factor of two or more.

Close interactions and mergers represent the most critical stage of galaxy evolution over cosmic time. Both models and observations indicate that, after the collision of a pair of massive, gas-rich spirals, the kinematic and structural properties of the remnant are consistent with elliptical galaxies (e.g., Barnes 1988; Dasyra et al. 2006). The relation between mergers and hot gaseous halos, instead, is still rather controversial. The first encounter of the progenitor disks may lead to substantial heating, through the dissipation of the kinetic energy in the gas components and the formation of copious shocks in the contact layers. In the present case, the former mechanism is largely insufficient. Assuming two identical progenitors, the amount of kinetic energy deposited during the merger is roughly M_gv²_c/8, where M_g is the mass of the X-ray emitting gas and v_c is the relative speed during the collision. The thermal energy content of the halo (E_th ≃ 5 × 10⁵⁸ erg) corresponds to v_c ∼ 1200 km s⁻¹, which is nearly twice the velocity observed in the Taffy Galaxies (Braine et al. 2003), a characteristic example of a head-on collision.¹⁸

On the other hand, the same halo luminosity is probably too large for the interaction of two galaxies with properties similar to the Milky Way. The peak of the X-ray luminosity in galaxy mergers is attained with the final coalescence, or shortly before (Brassington et al. 2007). According to simulations, its value depends on several factors, among which are the size and the gas fraction of the progenitors, the orbital parameters, and the relative orientation. The overall complexity is a possible reason for which extended X-ray emission is not systematically found in mergers. In principle, the presence of a supermassive black hole also has an important role in the gas heating and its subsequent dispersion (Cox et al. 2006). In NGC 6240, the AGN pair is believed to contribute significantly to the bolometric luminosity (Lutz et al. 2003; Egami et al. 2006), yet their direct feedback on the circumnuclear environment is difficult to establish. Their impact has presumably been negligible so far, as both nuclei are still completely enshrouded by a thick shell of dust and gas (Iwasawa & Comastri 1998; Vignati et al. 1999). In this view, unless the aforementioned physical parameters are unusually fine-tuned, a maximum L_X > 10⁴¹ erg s⁻¹ can be reached only if the progenitors are much more massive than the Milky Way (see also below). This makes of NGC 6240 an exceptional case study with respect to both the observations of known systems and the numerical simulations of mergers. Accepting that NGC 6240 is now at its peak X-ray luminosity, in fact, it should be kept in mind that the entire, diffuse X-ray emission is ∼3 times that of the halo alone.

The tidal forces can be responsible for the presence of hot gas at large distance from the center. The arms of the wide, cross-like morphological feature are openly suggestive of the typical tidal tails observed in mergers, but they have no detected counterpart at different wavelengths. Another problematic point resides in the gas temperature. When swept by an adiabatic shock, the gas is heated to kT_s = 3μm_pv²_s/16, where v_s is the speed of the shock front (Hollenbach & McKee 1979). The velocity required to produce a temperature of 0.65 keV is v_s ∼ 750 km s⁻¹, which is still too high when compared with the inferred orbital velocity of the two nuclei of NGC 6240 (155 km s⁻¹; Tecza et al. 2000), even if the rotation of the parent disks is taken into account. Finally, the inelastic nature of the galaxy merger does not seem capable of accounting for the energetics of the X-ray halo. Yet, the ongoing nuclear starburst is itself triggered by the global redistribution of the gas components that follows the strong gravitational disturbances, and represents a much more efficient energy source.

During a starburst, the formation of young, massive stars persists until the depletion of the molecular gas reservoir and/or the onset of some self-regulation mechanism. This activity leads to intense mass loss, stellar winds and frequent supernova (SN) explosions. The resultant deposition of energy, momentum, and metals into the intergalactic medium is a key driver of galaxy evolution. Based on the stellar K-band emission of NGC 6240, Tecza et al. (2000) evinced that a short (∼5 Myr) burst of star formation took place ∼20 Myr ago, possibly coincident with the latest perigalactic passage (but see Engel et al. 2010a). As the mechanical energy injected by a starburst is roughly 1% of its bolometric luminosity (e.g., Leitherer et al. 1999), and the total luminosity of NGC 6240 is L_bol ≈ L_IR ∼ 3 × 10⁴⁵ erg s⁻¹ (Armus et al. 2006), it takes at least ∼50 Myr for the starburst to supply the estimated thermal energy content of the X-ray halo, neglecting the AGN contribution to L_IR. It is therefore highly questionable whether the halo is a direct consequence of the most recent starburst episode, also considering its huge size, which calls for an average outflow velocity of ∼2500 km s⁻¹ over the last 20 Myr. This actually would be consistent with the fastest shocks we identified in the central 5 kpc (Wang et al. 2013), so the key issue is whether an outflow could sustain a similar expansion rate for several tens of kpc.

In the classical model of starburst-driven winds (Chevalier & Clegg 1985), the terminal outward velocity is $v_\infty = (2 \dot{E}/\dot{M})^{1/2}$, where $\dot{E}$ and $\dot{M}$ are the energy and mass injection rates, respectively. By adopting a standard mechanical energy input of 10⁵¹ erg per SN (Chevalier 1977), and an average SN mass of ∼10 M_☉ with intrinsic mass deposition fraction of ∼10%, we can write v_∞ ∼ 10⁴(ξ/Λ)^1/2 km s⁻¹, where the thermalization efficiency ξ accounts for any radiative losses, and the mass-loading factor Λ is the ratio between the total mass of the gas heated within the starburst and the mass of the direct SN ejecta (Veilleux et al. 2005). The obvious effect of radiative losses and mass loading is to slow down the wind. In the absence of observational constraints for ξ and Λ, the previous expressions can be rearranged further into the fully equivalent v_∞ = (5kT_c/μm_p)^1/2, which brings out the dependence on the central gas temperature only (Strickland & Heckman 2009). By approximating T_c with the nearly constant halo temperature, we obtain that v_∞ ≃ 720 km s⁻¹, commensurate with the average width of the optical line-emitting gas (Heckman et al. 1990), and with the terminal velocity of the nuclear outflow determined from the blueshifted Na i λλ5890, 5896 doublet (Na D) absorption (Heckman et al. 2000).

Alternatively, we assume that ξ ≃ 1, and attempt to assess Λ for a constant mass injection rate since the putative epoch of the most recent starburst. The SN rate in NGC 6240, as derived from the non-thermal radio continuum emission, is ∼2 yr⁻¹ (van der Werf et al. 1993; Beswick et al. 2001). This value strongly depends on the star formation history adopted (continuous, instantaneous, or merger induced), and is possibly overestimated by up to an order of magnitude (Engel et al. 2010a). At least, it represents a useful upper limit for the present purpose.¹⁹ A total mass of hot gas in the halo of ∼10¹⁰ M_☉ (Table 6) indicates that Λ ∼ 300. Taken at the face value, this would imply a mass-loading efficiency (i.e., the mass outflow rate normalized to the star formation rate) larger than 1, which is not unusual among ULIRGs (Rupke et al. 2005), and is comparable to that inferred for similar IR-luminous, merging systems like Mrk 266 (Wang et al. 1997) and Arp 299 (Heckman et al. 1999). In both of the latter cases, however, the dynamical age of the X-ray nebula is broadly consistent with the typical starburst lifetime. In NGC 6240, such a substantial mass loading would correspond to v_∞ ≃ 700 km s⁻¹, in excellent agreement with our previous estimate. In conclusion, any galactic wind originating some 20 Myr ago must have traveled for ∼10–15 kpc at most. Remarkably, this is just beyond the size of the optical nebula and of the soft X-ray core.

The two equivalent expressions for the terminal velocity used above are derived from wind models that neglect both ambient gas and gravitational forces. The mass loading is centralized (i.e., it takes place within the starburst region itself), and is typically limited to a few M_☉ per SN (Λ < 10; Suchkov et al. 1996). The archetypal source to which these analytical models and hydrodynamical simulations are compared is M82, since it is isolated, contains no AGN, and the hot gas is free to escape perpendicular to the galactic plane (Griffiths et al. 2000). The huge mass loading required for mergers, instead, implies the entrainment of tens to hundreds of M_☉ per SN in the ambient material swept up by the wind. In a ULIRG-like system, due to the high density and the disturbed geometry, it is plausible that the conversion of mechanical energy into radiation is much less efficient than in M82-like starbursts (e.g., Colina et al. 2004), and even that outflows expand only over short distances before terminating. However, as the clumpy morphological structures in NGC 6240 reach out to the largest physical scales, a wind nature for the entire halo cannot be conclusively rejected. Enhanced star formation with respect to quiescent galaxies has been likely in place since the beginning of the interaction, dated to several hundreds of Myr ago. Moreover, in these earlier times the additional momentum injection due to the AGN radiation pressure could have been non-negligible, with a possible boost effect on the starburst outflows.

The secular wind scenario, in keeping with the dynamical age of ∼200 Myr, can be probed through the average halo properties. In particular, the gas density in a freely expanding wind should decrease as ∼r⁻², as a consequence of the mass conservation law with constant outflow velocity at large distances. If the gas is nearly isothermal, the observed luminosity is related to the emission measure in a straightforward manner, hence the surface brightness is expected to exhibit a power-law dependence on radius as well: qualitatively, Σ∝n²_edV/dS ∼ r⁻³. Incidentally, this behavior was first observed in M82 with Einstein (Fabbiano 1988). We have then fitted the radial profile of Figure 3 at distances r > 12 kpc, to avoid any contamination from the core component. As already indicated by the large scale radius in the β-model fit (Table 2), a power-law shape gives a very poor description (χ²/ν ∼ 171/24) of the average surface brightness. Indeed, over the range explored, Σ(r) follows very well (χ²/ν ≃ 16/24) an exponential decline with scale length l = 10.1(± 0.2) kpc. The temperature and density profiles are plotted in Figure 15. Although only five measurements are available from the spectral fits to regions H1–H5, and the relative uncertainties are quite large, the halo gas density decreases with radius as n_e∝r^−α with α ≃ 0.75(± 0.15), showing a milder trend with respect to a single adiabatic flow. The temperature is consistent with being constant beyond ∼5 kpc. These features could be explained by the superposition of successive winds emanating from different locations, suggesting a widespread rather than centrally concentrated starburst.

Zoom In Zoom Out Reset image size

For completeness, we have extended the kT and n_e profiles to smaller radii by extracting the spectra from three annular regions in the core, which provides twice as many counts as the halo over the 0.5–1.5 keV energy range. In this case, our single thermal model is purely phenomenological, since a multi-temperature plasma is known to be present in the nuclear environment (Netzer et al. 2005; Wang et al. 2013). This was partly compensated through the power-law component, whose photon index was left free to vary and constrained up to 5 keV. Moderate column densities (a few ×10²¹ cm⁻² around the nuclei, then reducing with distance) were also required. In brief, within the inner regions the gas density becomes compatible with the r⁻² trend, while a central temperature T_c ∼ 1–1.5 keV does not substantially modify the wind expansion range suggested above. This simplified, preliminary analysis confirms that the soft X-ray emission in the core of NGC 6240, which is spatially correlated with the Hα filaments, is most likely dominated by shock-heated material in a recent starburst-driven wind, whose connection with the simultaneous black hole growth (e.g., King 2005) is hard to determine. As the detailed study of this component is the subject of a companion paper (J. Wang et al. 2013, in preparation), its physical properties are not discussed further in this context.

The large deviations from a classical wind model, in particular the flat temperature distribution, hint at near-isothermal, hydrostatic equilibrium. Together with the long cooling time (a few Gyr), this implies some form of gravitational confinement. Given the extent of the hot-gas halo, the baryonic matter is likely just a fraction of the total gravitating mass. Assuming the existence of a dark matter halo with a Navarro–Frenk–White density profile (Navarro et al. 1997), the gas virial temperature can be written as follows:

$$kT_{\mathrm{vir}} = \frac{1}{2} \mu m_p \frac{GM_{\mathrm{vir}}}{r_{\mathrm{vir}}} \simeq 0.135 \left(\!\frac{M_{\mathrm{vir}}}{10^{12}\, M_{\odot }} \!\right) \left(\!\frac{r_{\mathrm{vir}}}{100\, {\mathrm{kpc}}} \!\right)^{-1} \,{\mathrm{keV}},$$

where r_vir is the virial radius, usually taken to enclose an overdensity by a factor of 200 with respect to either the mean or the critical cosmic value at the time of the gravitational collapse, and M_vir the virial mass within the corresponding volume, serving as a proxy for the total halo mass. Accordingly, the dark halo of NGC 6240 has to be both massive and compact in order to confine the soft X-ray emitting gas within its potential well. Making use of the relations in Mo & White (2002) we can render explicit the dependence of r_vir on mass and redshift, obtaining that kT_vir ∼ 0.042M^2/3₁₂(1 + z) keV, where M₁₂ is the virial mass expressed in units of 10¹² M_☉.

A possible evolutionary scenario consistent with the observed gas temperature of 0.65 keV is that of a cold dark matter halo formed at z ∼ 2, with a total mass of ∼10¹³ M_☉. This value marks the standard separation between galaxy-scale and group-scale halos (see Humphrey et al. 2006). For comparison, the Milky Way is estimated to have a virial mass of ∼1.3 × 10¹² M_☉ (McMillan 2011). NGC 6240 is then consistent with the central remnant of a group of galaxies. This interpretation was first proposed by Huo et al. (2004), who considered the huge content of hot gas (>10¹⁰ M_☉), the low metallicity of the outer halo, and the large velocity dispersion (the full width at zero intensity of CO emission lines reaches up to ∼1400 km s⁻¹; Feruglio et al. 2013). However, no evidence has been reported so far of a present-day galaxy overdensity in the surroundings of NGC 6240. Any other group member may have experienced a merger in the past, but in this case the progenitors of the final coalescence we are now witnessing would be elliptical-like, or highly irregular, rather than disk-like.

The degree of the uncertainties involved in linking the spectroscopic and virial temperatures (e.g., Ciotti & Pellegrini 2008), and some ambiguity in the definitions themselves (e.g., Voit 2005), should also be kept in mind. Moreover, the gas temperature estimated through the single thermal model is only a first approximation, and could be somewhat unreliable. The best fit to the FH spectrum calls for two components with different temperatures and abundances (model E). It is then reasonable to associate the warmer (kT₁ ∼ 0.8 keV) and more metal-rich (Z_α ∼ 0.5 solar) component with the chemically evolved, starburst-injected gas, and to identify the cooler (kT₂ ∼ 0.25 keV) and more metal-poor (Z ≡ 0.1 solar) component with the gravitationally bound, pre-existing halo material. Also the energy dissipated during the merger process may contribute to the heating of the ambient gas. In this view, a fossil group nature for the system would not be strictly required, and the dark matter halo of NGC 6240 would be broadly consistent with just a pair of massive spiral galaxies.

The lower temperature of ∼0.25 keV would also be in good agreement with the virial temperature kT_σ ∼ μm_pσ²_* ∼ 0.3 keV derived from the average stellar velocity dispersion, σ_* ≃ 200–220 km s⁻¹ (Engel et al. 2010a). This comparison is physically motivated only for early-type galaxies (e.g., Pellegrini 2011), which are actually believed to represent the ultimate evolutionary stage of a major merger such as NGC 6240. Although the potential of the nuclear dynamical mass is expected not to be dominant on the halo scales, the stellar radial surface brightness profile follows the r^1/4 law typical of elliptical galaxies out to ∼20–25 kpc (Bush et al. 2008). While some of the disk-like structures are still present, the whole remnant has apparently entered the final relaxation phase.

Another primary source of information to understand the nature of the X-ray halo is the metallicity pattern, including absolute and relative abundances and their dependence on galactocentric distance. The average α-element to iron abundance ratio turns out to be generally supersolar all across the halo (Table 7). This is a key observable for tracing the chemical evolution of the interstellar matter, due to the different production yields and typical timescales of the various SNe populations involved. Synthesis models for type II SNe (i.e., core-collapsed massive stars) predict Si/Fe ratios up to ∼3–5 solar (e.g., Nomoto et al. 2006), while these drop to ∼0.5 solar for type Ia SNe (i.e., exploded white dwarfs in close binary systems). Evidence for a supersolar Si/Fe ratio, for instance, was found in galaxy mergers like the Antennae (Baldi et al. 2006a, 2006b), and has been recently revealed in the central regions of young elliptical galaxies, which are the sites of the latest (a few tens of Myr), merger-induced star formation (Kim et al. 2012).

In Figure 16, the abundance ratios Z_Mg/Z_Fe and Z_Si/Z_Fe are plotted against each other as diagnostics of the enrichment history in the six azimuthal sectors of the halo. For comparison, the same values are shown for different regions of the disturbed spiral galaxy NGC 4490, which is interacting with the nearby irregular companion NGC 4485 (Richings et al. 2010), and of the elliptical galaxies analyzed by Kim et al. (2012). In such a diagram, the theoretical yields from various models of type Ia SNe (Nomoto et al. 1997) and type II SNe (Nagataki & Sato 1998) have a wide separation. All of the S1–S6 halo subregions definitely occupy the typical location of starburst environments, dominated by a young stellar population and type II SNe. From an evolutionary perspective, NGC 6240 should gradually move toward the intermediate position of early-type galaxies, following the takeover by type Ia SNe after the merger completion and the starburst fading.

Zoom In Zoom Out Reset image size

There is actually a possible caveat to the picture outlined above. By assuming in our fits a solar abundance for nickel, we obtain a nominal Ni/Fe ratio ∼7 times above solar in the full halo. Similar findings have been discussed over the past two/three decades for SN remnants (based on the strength of forbidden optical/near-IR [Ni ii] lines; e.g., Bautista et al. 1996) and galaxy clusters (e.g., Dupke & White 2000). On average, type Ia SNe are expected to have a much higher production yield for nickel with respect to type II SNe. A large relative abundance of nickel, if real, would imply a more complex enrichment history. As mentioned before, however, Z_Ni can only be poorly constrained, due to its degeneracy with Z_Ne. Our assumption of a single thermal component might also introduce some bias. Indeed, by tying nickel and iron abundances in the two-temperature model D we achieve a statistically equivalent fit, with Δχ² < 0.5, and just an increase of Z_Ne to ∼0.6 solar. In general, the best-fitting abundance ratios from model D are slightly lower (Table 7), but still in excellent agreement with the predicted yields of type II SNe for all elements, nickel included.

Based on the above considerations, the whole halo of NGC 6240 appears to have experienced a significant metal enrichment by type II SNe. The constant abundance ratios in the inner and outer halo regions imply that the enrichment process has been uniform. As a consequence, the gas redistribution cannot be associated with recent starburst episodes, but has rather proceeded in the whole course of the interaction, whose timescale up to now is presumably ∼1 Gyr. Mergers of identical disk galaxies with massive stellar bulges are characterized by a single, intense peak of star formation during the nuclear coalescence (Springel et al. 2005; Cox et al. 2008). Entering this major burst, NGC 6240 is going to exceed the ULIRG luminosity threshold (Tacconi et al. 1999). The final stage is not imminent yet, since the two nuclei (which are the remnant of the progenitors' bulges) still have a projected separation of ∼0.8 kpc (15; Max et al. 2007) and are possibly in the process of separating again (Tecza et al. 2000; Engel et al. 2010a). This notwithstanding, the identification of several tens of young circumnuclear star clusters, with ages in the range ∼5–15 Myr, confirms that enhanced star formation activity compared to isolated, quiescent galaxies is currently ongoing (Pasquali et al. 2003; Pollack et al. 2007). A steady star formation rate of this kind (61 M_☉ yr⁻¹; Yun & Carilli 2002) over the past ∼200 Myr (the dynamical age of the soft X-ray halo), perhaps involving the peripherical regions, can easily account for the energy and metallicity content of the hot gas component. The tentative decrease with radius of the absolute abundances hints at the gentle mixing of the starburst-driven outflows with a pre-existing halo medium, as also suggested by the best-fitting spectral model E. If no dilution effect were present and the dispersion of metal-rich gas were highly efficient, the observed metallicity gradient would be positive, with supersolar abundances at the larger distances (Cox et al. 2006).

The fate of the starburst-processed gas is not clear. Some merger remnants, and young ellipticals bearing the signatures of recent interactions, are under-luminous in the X-rays with respect to relaxed early-type galaxies (Fabbiano & Schweizer 1995; Read & Ponman 1998), while different mechanisms of halo regeneration have been proposed (O'Sullivan et al. 2001a). If the wind temperature is much larger than the virial temperature, most of the hot gas is not bound and will eventually escape the system, becoming undetectable in the X-rays due to the density drop. In NGC 6240, the single-temperature cooling time of a few Gyr implies that the halo luminosity will not be fully extinguished after the merger conclusion. In the more complex scenarios, the different gas phases might have much shorter cooling times, depending on their degree of mixing. According to model E, unless the filling factors are very small (η ≪ 0.1), the radiative cooling of the warmer gas component will take several hundreds of Myr at least. Provided that some favorable conditions are met (e.g., the ultimate AGN feedback is not dramatically powerful, the cooler gas component exerts a sufficient drag, and the gravitational potential well is deep enough), NGC 6240 could be a case of major merger with no complete blowout.